Constructing imaginary discourse trees to improve answering convergent questions

Abstract

Systems and methods for improving question-answering recall for complex, multi-sentence, convergent questions. More specifically, an autonomous agent accesses an initial answer that partly answers a question received from a user device. The agent represents the question and the initial answer as discourse trees. From the discourse trees, the agent identifies entities in the question that are not addressed by the answer. The agent forms an additional discourse tree from an additional resource such as a corpus of text. The additional discourse tree rhetorically connects a non-addressed entity with the answer. The agent designates this discourse tree as an imaginary discourse tree. When combined with the initial answer discourse tree, the imaginary discourse tree is used to generate an improved answer relative to existing solutions.

What is claimed is: 
     
       1. A method comprising:
 constructing, using a computing device and from a question, a question discourse tree comprising a question entity, wherein the question discourse tree represents rhetorical relationships between elementary discourse units of the question; 
 accessing, using the computing device and from a corpus of text, an initial answer; 
 constructing, using the computing device, from the initial answer, an answer discourse tree comprising an answer entity, wherein the answer discourse tree represents rhetorical relationships between elementary discourse units of the initial answer; 
 determining, using the computing device, that a score indicating a relevance of the answer entity to the question entity is below a threshold; 
 generating an imaginary discourse tree by:
 creating, from the corpus of text, an additional discourse tree; 
 determining that the additional discourse tree comprises a rhetorical relation that connects the question entity with the answer entity; 
 extracting a sub-tree of the additional discourse tree comprising the question entity, the answer entity, and the rhetorical relation, thereby generating an imaginary discourse tree; and 
 
 outputting an answer represented by a combination of the answer discourse tree and the imaginary discourse tree. 
 
     
     
       2. The method of  claim 1 , wherein accessing the initial answer comprises:
 determining an answer relevance score for a portion of text; and 
 responsive to determining that the answer relevance score is greater than a threshold, selecting the portion of text as the initial answer. 
 
     
     
       3. The method of  claim 1 , wherein the imaginary discourse tree comprises a node representing the rhetorical relation, the method further comprising integrating the imaginary discourse tree into the answer discourse tree by connecting the node to the answer entity. 
     
     
       4. The method of  claim 1 , wherein creating the additional discourse tree comprises:
 calculating, for each of a plurality of additional discourse trees, a score that indicates a number of question entities that comprise a mapping to one or more answer entities in the respective additional discourse tree; and 
 selecting, from the plurality of additional discourse trees, an additional discourse tree with a highest score. 
 
     
     
       5. The method of  claim 1 , wherein creating the additional discourse tree comprises:
 calculating, for each of a plurality of additional discourse trees, a score by applying a trained classification model to one or more of (a) the question discourse tree and (b) the respective additional answer discourse tree; and 
 selecting, from the plurality of additional discourse trees, an additional discourse tree with a highest score. 
 
     
     
       6. The method of  claim 1 , wherein the question comprises a plurality of keywords, and wherein accessing the initial answer comprises:
 obtaining a plurality of answers based on a search query comprising the keywords by performing a search of a plurality of electronic documents; 
 determining, for each of the plurality of answers, an answer score indicating a level of match between the question and the respective answer; and 
 selecting, from the plurality of answers, an answer having a highest score as the initial answer. 
 
     
     
       7. The method of  claim 1 , wherein calculating the score comprises:
 applying a trained classification model to one or more of (a) the question discourse tree and (b) the answer discourse tree; and 
 receiving the score from the classification model. 
 
     
     
       8. The method of  claim 1 , wherein constructing a discourse tree comprises:
 accessing a sentence comprising a plurality of fragments, wherein at least one fragment comprises a verb and a plurality of words, each word comprising a role of the words within the fragment, wherein each fragment is an elementary discourse unit; and 
 generating a discourse tree that represents rhetorical relationships between the plurality of fragments, wherein the discourse tree comprises a plurality of nodes, each nonterminal node representing a rhetorical relationship between two of the plurality of fragments, each terminal node of the nodes of the discourse tree is associated with one of the plurality of fragments. 
 
     
     
       9. The method of  claim 1 , further comprising:
 determining, from the question discourse tree, a question communicative discourse tree comprising a question root node, wherein a communicative discourse tree is a discourse tree that includes communicative actions, and wherein the generating further comprises: 
 determining, from the an imaginary discourse tree, an answer communicative discourse tree, wherein the answer communicative discourse tree comprises an answer root node; 
 merging the communicative discourse trees by identifying that the question root node and the answer root node are identical; 
 computing a level of complementarity between the question communicative discourse tree and the answer communicative discourse tree by applying a predictive model to the merged communicative discourse tree; and 
 responsive to determining that the level of complementarity is above a threshold, outputting a final answer corresponding to the imaginary discourse tree. 
 
     
     
       10. The method of  claim 1 , wherein a discourse tree represents rhetorical relationships between a plurality of fragments of text, wherein the discourse tree comprises a plurality of nodes, each nonterminal node representing a rhetorical relationship between two of the plurality of fragments, each terminal node of the nodes of the discourse tree is associated with one of the plurality of fragments; and wherein constructing a communicative discourse tree comprises:
 matching each fragment that has a verb to a verb signature by:
 accessing a plurality of verb signatures, wherein each verb signature comprises the verb of the fragment and a sequence of thematic roles, wherein thematic roles describe the relationship between the verb and related words; 
 determining, for each verb signature of the plurality of verb signatures, a plurality of thematic roles of the respective signature that match a role of a word in the fragment; 
 selecting a particular verb signature from the plurality of verb signatures based on the particular verb signature comprising a highest number of matches; and 
 associating the particular verb signature with the fragment. 
 
 
     
     
       11. A computer-implemented method comprising:
 constructing, for a question, a question discourse tree comprising a plurality of question entities; 
 constructing, from an initial answer, an answer discourse tree comprising a plurality of answer entities; 
 establishing, between a first question entity of the plurality of question entities and an answer entity of the plurality of answer entities, a mapping that establishes a relevance of the answer entity to the first question entity; 
 responsive to determining that a second question entity of the plurality of question entities is not addressed by any of the plurality of answer entities, generating an imaginary discourse tree by combining an additional discourse tree corresponding to an additional answer with the answer discourse tree; 
 determining, from the question discourse tree, a question communicative discourse tree; 
 determining, from the an imaginary discourse tree, an answer communicative discourse tree; 
 computing a level of complementarity between the question communicative discourse tree and the answer communicative discourse tree by applying a predictive model to the question communicative discourse tree and the answer communicative discourse tree; and 
 responsive to determining that the level of complementarity is above a threshold, outputting a final answer corresponding to the imaginary discourse tree. 
 
     
     
       12. A system comprising:
 a non-transitory computer-readable medium storing computer-executable program instructions; and 
 a processing device communicatively coupled to the non-transitory computer-readable medium for executing the computer-executable program instructions, wherein executing the computer-executable program instructions configures the processing device to perform operations comprising:
 constructing, using a computing device and from a question, a question discourse tree comprising a question entity, wherein the question discourse tree represents rhetorical relationships between elementary discourse units of the question; 
 accessing, using the computing device and from a corpus of text, an initial answer; 
 constructing, using the computing device, from the initial answer, an answer discourse tree comprising an answer entity, wherein the answer discourse tree represents rhetorical relationships between elementary discourse units of the initial answer; 
 determining, using the computing device, that a score indicating a relevance of the answer entity to the question entity is below a threshold; 
 generating an imaginary discourse tree by:
 creating, from the corpus of text, an additional discourse tree; 
 determining that the additional discourse tree comprises a rhetorical relation that connects the question entity with the answer entity; 
 extracting a sub-tree of the additional discourse tree comprising the question entity, the answer entity, and the rhetorical relation, thereby generating an imaginary discourse tree; and 
 
 outputting an answer represented by a combination of the answer discourse tree and the imaginary discourse tree. 
 
 
     
     
       13. The system of  claim 12 , wherein accessing the initial answer comprises:
 determining an answer relevance score for a portion of text; and 
 responsive to determining that the answer relevance score is greater than a threshold, selecting the portion of text as the initial answer. 
 
     
     
       14. The system of  claim 12 , wherein the imaginary discourse tree comprises a node representing the rhetorical relation, the operations further comprising integrating the imaginary discourse tree into the answer discourse tree by connecting the node to the answer entity. 
     
     
       15. The system of  claim 12 , wherein creating the additional discourse tree comprises:
 calculating, for each of a plurality of additional discourse trees, a score that indicates a number of question entities that comprise a mapping to one or more answer entities in the respective additional discourse tree; and 
 selecting, from the plurality of additional discourse trees, an additional discourse tree with a highest score. 
 
     
     
       16. The system of  claim 12 , wherein creating the additional discourse tree comprises:
 calculating, for each of a plurality of additional discourse trees, a score by applying a trained classification model to one or more of (a) the question discourse tree and (b) the respective additional answer discourse tree; and 
 selecting, from the plurality of additional discourse trees, an additional discourse tree with a highest score. 
 
     
     
       17. The system of  claim 12 , wherein the question comprises a plurality of keywords, and wherein accessing the initial answer comprises:
 obtaining a plurality of answers based on a search query comprising the keywords by performing a search of a plurality of electronic documents; 
 determining, for each of the plurality of answers, an answer score indicating a level of match between the question and the respective answer; and 
 selecting, from the plurality of answers, an answer having a highest score as the initial answer. 
 
     
     
       18. The system of  claim 12 , wherein calculating the score comprises:
 applying a trained classification model to one or more of (a) the question discourse tree and (b) the answer discourse tree; and 
 receiving the score from the classification model. 
 
     
     
       19. The system of  claim 12 , wherein constructing a discourse tree comprises:
 accessing a sentence comprising a plurality of fragments, wherein at least one fragment comprises a verb and a plurality of words, each word comprising a role of the words within the fragment, wherein each fragment is an elementary discourse unit; and 
 generating a discourse tree that represents rhetorical relationships between the plurality of fragments, wherein the discourse tree comprises a plurality of nodes, each nonterminal node representing a rhetorical relationship between two of the plurality of fragments, each terminal node of the nodes of the discourse tree is associated with one of the plurality of fragments. 
 
     
     
       20. The system of  claim 12 , wherein executing computer-executable program instructions configures the processing device to perform operations comprising:
 determining, from the question discourse tree, a question communicative discourse tree comprising a question root node, wherein a communicative discourse tree is a discourse tree that includes communicative actions, and wherein the generating further comprises: 
 determining, from the an imaginary discourse tree, an answer communicative discourse tree, wherein the answer communicative discourse tree comprises an answer root node; 
 merging the communicative discourse trees by identifying that the question root node and the answer root node are identical; 
 computing a level of complementarity between the question communicative discourse tree and the answer communicative discourse tree by applying a predictive model to the merged communicative discourse tree; and 
 responsive to determining that the level of complementarity is above a threshold, outputting a final answer corresponding to the imaginary discourse tree.